"What should my similarity score be?" is the most-asked — and most-misunderstood — question in academic writing. The short answer: there is no single correct number, and reducing the question to a number leads you astray. What matters is not the percentage in the report but what that percentage represents. This guide shows you how to read a similarity score, which thresholds are used in practice, and how to close the risk before you submit.
Is the similarity score the same as plagiarism?
No — and that distinction is the foundation of everything. What tools like Turnitin and iThenticate produce is a similarity score: it shows how much your text overlaps, word for word, with other sources in the tool's database. That score is not proof of plagiarism. Properly cited direct quotations, methods sentences that are standard in your field ("data were analyzed using SPSS..."), and your reference list all add to it. So even a clean, honest paper has a non-zero similarity score. Think of the score as a scanner: it tells you where to look; the editor makes the call.
What is an acceptable similarity score?
The threshold is always decided by the target journal or institution; different bodies set different limits. Still, common practice clusters around these bands:
| Overall similarity | What it means in practice |
|---|---|
| 0-10% | A comfortable zone for most journals and institutions; usually no action needed. |
| 10-20% | The most common upper-bound range. Most journals accept this band — but still check the matches in the report. |
| 20-30% | Risky. Some institutions tolerate this much (especially in theses, excluding the bibliography); most journals will scrutinize it. |
| 30%+ | An automatic reject or a serious revision request in most places; review the content from scratch. |
These bands are a reference, not a hard rule. The only binding number is the one in your target journal's author guidelines — always read it before you submit.
The key point: once the bibliography and direct quotations are excluded, the real figure is usually markedly lower than the raw report. So instead of panicking at a single percentage, open the report and look at where the matches come from.
Why does single-source similarity matter more than the total?
Two papers can both show 12% overall and be in completely different situations. In the first, the similarity is spread across dozens of sources, each a few words of standard phrasing — that is normal. In the second, if 10 of those 12 points come from one paper, that is a copied paragraph and a serious plagiarism signal, even though the overall figure is low. This is why experienced editors look first at the single-source match; most treat overlap above ~2-5% from one source as a red flag. When you read the report, sort by percentage and start with the highest single match.
What should a similarity report look like for indexing and tenure?
In Türkiye this is not only an ethical matter but a procedural requirement. A similarity report is now the standard, not the exception: 79% of the DOAJ-registered open-access journals in Peerfect's corpus state that they run plagiarism checks in their submission process (Academic Publishing Landscape 2026). TR Dizin criteria also require journals seeking inclusion to use a similarity (plagiarism) report in their editorial process; so if you submit to a TR Dizin-indexed journal, a report will be requested sooner or later. Likewise, for thesis submission and associate-professor (doçentlik) applications, most institutions require a Turnitin or iThenticate report. The practical takeaway: generate the report yourself, before submission, because the editor or committee will generate it anyway.
How do you lower the score before submission?
The goal is not to "game" the number but to make the text honestly clean:
- Put direct quotations in quotation marks and cite them. A properly attributed quote is not plagiarism; an unattributed copy is.
- Rewrite copied sentences in your own words — preserving the meaning, not just swapping synonyms. Genuine paraphrasing still cites the source.
- Trim standard phrasing. Boilerplate sentences in the methods and introduction inflate the score; cut what you don't need.
- Watch your own prior work (self-plagiarism): reusing your own text without citing it is still a problem.
- Avoid automated "reduction" with AI. Auto-paraphrase tools distort meaning and increasingly trip AI-content detectors — a separate ground for rejection. Keep the final say yourself.
How does high similarity lead to a desk reject?
A high or badly distributed similarity score is one of the classic reasons a paper gets desk-rejected before it ever reaches a reviewer. Before starting peer review, the editor checks the similarity report during the desk screen; a report over the limit is enough to eliminate the paper on day one. The good news: this is a completely preventable kind of rejection — because you can see the report before submission too.
Peerfect's free pre-review engine scans your paper through an editor's and reviewer's eyes before you submit; alongside plagiarism risk it surfaces the other desk-reject triggers — scope fit, formatting, methodology — in one pass. High similarity is not a problem in isolation: understanding why papers get rejected and choosing the right journal from the start are parts of the whole. Look at the nature of the match, not the number; the rest is fixable.